Current Progress
Week 1completed
Protector Mindset & Secure Development Environment
Objectives
- •Understand the AI Protector role within the changing-roles framework
- •Set up secure local development with Node.js, Git, VS Code, Copilot, and Claude Desktop
- •Document environment hardening steps including extensions, policies, and secrets management
- •Connect to existing MCP servers (Rolldice, Bootcamp Agent, Calendar Booking) and evaluate security baselines
- •Collect Australian security case studies and LMS security journey insights for contextual awareness
Lessons Learned
- →Security must be considered from the very first commit, not as an afterthought
- →Development environment security is foundational - compromised dev tools can lead to supply chain attacks
- →MCP servers introduce new attack surfaces that require specific security assessments
- →Australian data residency and privacy regulations (Privacy Act 1988) have specific implications for AI systems
- →Claude Desktop and VS Code Copilot have different security and privacy models that must be understood
Key Takeaways
- ✓Shift-left security starts with securing the development environment itself
- ✓Every MCP server connection should be evaluated for data handling, privacy, and security risks
- ✓Understanding local case studies helps contextualize threats relevant to Australian organizations
Challenges
- ⚠Balancing developer productivity with security constraints
- ⚠Understanding the full scope of data that MCP servers can access
- ⚠Identifying which security controls are appropriate for different development scenarios
Week 2in progress
Cyber Security Bootcamp Integration & Digital Portfolio Foundation
Objectives
- •Integrate LMS video content from Cyber Security Bootcamp into learning journey
- •Establish baseline security posture for digital portfolio application
- •Review OWASP Top 10 in the context of AI-powered web applications
- •Implement initial authentication and authorization patterns
- •Document security decisions and rationale for portfolio foundation
Lessons Learned
- →Traditional web security principles (OWASP Top 10) still apply but need AI-specific adaptations
- →Authentication for AI-powered applications requires considering both human and agent access patterns with Clerk.js
- →Digital portfolios can be high-value targets due to professional information and contact details
- →LMS integration provides structured learning path aligned with hands-on implementation
- →Security documentation from day one creates an audit trail and learning reference
- →Next.js 15 App Router provides enhanced security patterns with Server Components as secure-by-default
- →Vercel Analytics and Speed Insights integration adds performance monitoring without compromising security
Key Takeaways
- ✓Foundation security decisions have cascading effects throughout the application lifecycle
- ✓Combining theoretical LMS content with practical implementation reinforces learning
- ✓Security patterns for Next.js applications differ from traditional server-rendered apps
- ✓ClerkProvider setup in root layout enables role-based access control (RBAC) for admin routes
- ✓Resource hints and DNS prefetching improve performance while maintaining security boundaries
Challenges
- ⚠Determining appropriate security level for a public portfolio vs. enterprise application
- ⚠Integrating multiple authentication providers (Clerk, OAuth) cohesively
- ⚠Balancing public accessibility with protection of sensitive admin functions
Week 3completed
Secure the My Digital Portfolio Application
Objectives
- •Implement comprehensive security controls for Next.js portfolio application
- •Configure Clerk authentication with role-based access control (RBAC)
- •Secure API routes with proper authentication and rate limiting
- •Implement Content Security Policy (CSP) and other security headers
- •Secure database operations and prevent common injection attacks
- •Set up environment variable management and secrets handling
Lessons Learned
- →✅ IMPLEMENTED: Edge middleware with Clerk authentication enforces RBAC at runtime='experimental-edge'
- →✅ IMPLEMENTED: Arcjet SDK provides shield (XSS/SQL injection), bot detection, and token bucket rate limiting (20 tokens/10s, capacity 50)
- →✅ IMPLEMENTED: PostgreSQL with pg-vector extension using parameterized queries prevents SQL injection
- →✅ IMPLEMENTED: Clerk's publicMetadata.role='admin' enables flexible RBAC without custom auth tables
- →✅ IMPLEMENTED: Environment variables managed securely (DATABASE_URL, ARCJET_KEY, CLERK secrets) with .env.local
- →✅ IMPLEMENTED: Route matcher pattern protects /admin routes while allowing public access to /admin/sign-in
- →✅ IMPLEMENTED: Custom error handling provides user-friendly messages without leaking system details
- →✅ IMPLEMENTED: Connection pooling (max: 20 connections) with automatic client release prevents resource exhaustion
- →Cache-Control headers set for public API routes: 'public, s-maxage=300, stale-while-revalidate=600'
- →Performance optimization: DNS prefetching for api.openai.com, vercel.live, clerk.com reduces latency
Key Takeaways
- ✓Defense in depth achieved: Arcjet (application layer) + Clerk (authentication) + PostgreSQL (data layer) + Middleware (edge)
- ✓Edge middleware with experimental-edge runtime provides global security enforcement with minimal latency
- ✓Graceful degradation: Arcjet returns mock decision when not configured, enabling development without all keys
- ✓Security monitoring via Vercel Analytics, Toaster notifications (sonner), and database health checks
- ✓Suspense boundaries prevent security-sensitive content from streaming before auth checks complete
Challenges
- ⚠✅ SOLVED: CSP compatibility with Next.js - using Tailwind with custom properties and system theme detection
- ⚠✅ SOLVED: Rate limiting balance - Arcjet token bucket with refillRate: 20 allows normal usage while blocking abuse
- ⚠✅ SOLVED: Secret management - Vercel environment variables with separate dev/staging/production contexts
- ⚠✅ SOLVED: Admin route protection - createRouteMatcher with isAdminRoute and isPublicRoute patterns
- ⚠ONGOING: Performance vs security tradeoff - monitoring s-maxage cache effectiveness with Speed Insights
Upcoming Weeks
Week 4in progress
Layered Defenses: WAF, Arcjet, and Vercel Firewall
Objectives
- •Configure Vercel Firewall for edge protection
- •Integrate Arcjet SDK for AI-aware rate limiting and bot protection
- •Implement Web Application Firewall (WAF) rules
- •Set up DDoS protection and monitoring
- •Test and tune firewall rules with real traffic patterns
Lessons Learned
- →✅ IMPLEMENTED: Arcjet shield in 'LIVE' mode blocks SQL injection and XSS attacks at application layer
- →✅ IMPLEMENTED: Bot detection allows SEARCH_ENGINE, MONITOR, and PREVIEW categories while blocking malicious bots
- →✅ IMPLEMENTED: Token bucket rate limiting with characteristics=['ip.src'] prevents per-IP abuse
- →✅ IMPLEMENTED: checkArcjetProtection() helper function used across /api/mcp and /api/contact routes
- →Vercel Firewall provides edge-level DDoS protection and geographic filtering (regions: ['pdx1'])
- →Multi-layer defense: Vercel Edge → Arcjet Application → Clerk Auth → Database validates defense-in-depth principle
Key Takeaways
- ✓Arcjet's 'DRY_RUN' mode enables testing rules before enforcement in production
- ✓Bot detection must balance security (blocking scrapers) with functionality (allowing search engines)
- ✓Rate limit tuning requires analytics: 20 tokens/10s allows ~120 requests/minute per IP
- ✓Graceful degradation with mock decisions ensures development continues without all API keys
Challenges
- ⚠Tuning rate limits to prevent false positives for power users while blocking abuse
- ⚠Determining appropriate bot allow-list categories for SEO vs security tradeoff
- ⚠Testing firewall rules without blocking legitimate traffic during deployment
Week 5planned
Kali Linux Penetration Testing Sprint
Objectives
- •Set up Kali Linux environment for security testing
- •Perform reconnaissance and vulnerability scanning with nmap, nikto, and dirb
- •Test rate limits with automated tools (Apache Bench, siege)
- •Attempt SQL injection against hardened database endpoints using sqlmap
- •Test XSS protection with XSS Hunter and manual payload injection
- •Attempt authentication bypass and session hijacking attacks
- •Document findings in professional penetration testing report format
Challenges
- ⚠Establishing legal and ethical boundaries for penetration testing on live production site
- ⚠Distinguishing between legitimate security testing and actual attacks in logs
- ⚠Documenting remediation steps without exposing vulnerabilities publicly
Week 6in progress
Prerequisites for Agent Security Advanced
Objectives
- •Review OAuth 2.1 and OpenID Connect specifications
- •Understand MCP authentication requirements and security model
- •Prepare environment for secure MCP server development
- •Study mcp-auth-demo reference implementation
- •Analyze existing MCP tools in portfolio for security gaps
Lessons Learned
- →✅ IMPLEMENTED: MCP server with 15+ tools providing professional data access (get_experiences, lookup_skills, query_projects)
- →✅ IMPLEMENTED: MCP tools use semantic search with vector embeddings for intelligent content retrieval
- →✅ IMPLEMENTED: Arcjet protection applied to /api/mcp endpoint prevents abuse of MCP tools
- →MCP protocol enables Claude Desktop, ChatGPT Developer Mode, and VS Code Copilot to access portfolio data
- →Current MCP implementation lacks OAuth - tools are protected by Arcjet rate limiting only
- →Model Context Protocol standardizes how AI agents access external data sources and APIs
- →Vector search with pg-vector (1536-dimensional embeddings) powers semantic search across professional content
Key Takeaways
- ✓MCP security requires authentication, authorization, and audit logging for production deployment
- ✓Current MCP tools demonstrate capability but need OAuth 2.1 upgrade for enterprise security
- ✓Semantic search via vector embeddings enables natural language queries across professional content
- ✓MCP integration in portfolio showcases cutting-edge AI agent communication standards
Challenges
- ⚠Balancing open access for portfolio visitors with secure API access for MCP clients
- ⚠Implementing OAuth 2.1 without breaking existing Claude Desktop integration
- ⚠Determining appropriate scopes and permissions for different MCP tool categories
Week 7planned
Agent Security Advanced Phase 1 & MCP Auth Demo
Resources
Objectives
- •Implement OAuth 2.1 authentication for MCP servers
- •Integrate mcp-auth-demo patterns into existing MCP implementation
- •Create authorization server for MCP client authentication
- •Secure MCP server endpoints with token validation
- •Test authentication flows with Claude Desktop and other MCP clients
- •Implement scope-based access control for different MCP tools
Challenges
- ⚠Maintaining backward compatibility with existing non-authenticated MCP clients
- ⚠Implementing token refresh flows for long-running AI agent sessions
- ⚠Debugging OAuth flows across multiple MCP client platforms
Week 8planned
Agent Security Advanced Phase 2 & Production Hardening
Objectives
- •Complete Agent Security Advanced curriculum
- •Implement production security hardening
- •Set up monitoring and alerting
- •Create incident response procedures
Week 9planned
Portfolio Integration & Security Reporting
Objectives
- •Integrate security features into portfolio showcase
- •Create executive security reports
- •Document security architecture and decisions
- •Prepare compliance documentation
Week 10planned
Final Presentation & Protector Launch Plan
Objectives
- •Complete final portfolio presentation
- •Deliver security playbooks and runbooks
- •Present executive briefing
- •Launch as AI Protector specialist
Implemented Security Features in This Portfolio
Real-world security implementations demonstrating AI Protector concepts
Authentication & Authorization
- Clerk.js Integration: Enterprise authentication with RBAC using publicMetadata.role
- Edge Middleware: clerkMiddleware with experimental-edge runtime for global auth enforcement
- Route Protection: createRouteMatcher pattern protects /admin/* routes dynamically
Application Security (Arcjet)
- Shield Protection: LIVE mode blocks SQL injection, XSS, and common web attacks
- Bot Detection: Allows search engines/monitors while blocking malicious bots
- Rate Limiting: Token bucket (20/10s refill, 50 capacity) per IP address
Database Security
- PostgreSQL + pg-vector: Parameterized queries prevent SQL injection
- Connection Pooling: Max 20 connections, 30s idle timeout, automatic release
- Transaction Safety: BEGIN/COMMIT/ROLLBACK with automatic error handling
MCP Security & AI Integration
- MCP Protocol: Standardized AI agent access with 15+ professional data tools
- Arcjet on /api/mcp: Rate limiting and shield protection for MCP endpoints
- OAuth 2.1 (Planned): Week 7-8 implementation for enterprise MCP security
Monitoring & Performance
- Vercel Analytics: Real-time traffic monitoring and performance metrics
- Speed Insights: Core Web Vitals tracking for performance optimization
- Health Checks: Database connectivity monitoring with response time alerts
Deployment Security
- Environment Isolation: Separate dev/staging/production secrets management
- Vercel Edge Network: Global CDN with DDoS protection (region: pdx1)
- SSL/TLS: Automatic HTTPS with production SSL certificate on custom domain
🎯 Security Metrics Summary
Rate Limit
20 req/10s
DB Pool
20 max conn
Cache TTL
300s + SWR
Edge Runtime
<50ms global
Overall Learning Insights
Key themes emerging from the AI Protector journey
🛡️ Security as a Mindset
The AI Protector role is not just about implementing security controls, but about developing a security-first mindset that questions assumptions, considers threat models, and evaluates risks at every stage of development. This portfolio demonstrates security integrated from the first commit.
🔐 Layered Defense Strategy (Defense in Depth)
No single security control is sufficient. This project implements defense in depth with multiple overlapping controls:
- Edge Layer: Vercel Firewall with DDoS protection and geographic filtering
- Application Layer: Arcjet SDK with shield (XSS/SQL injection), bot detection, and rate limiting
- Authentication Layer: Clerk.js with RBAC enforced via edge middleware
- Data Layer: PostgreSQL with parameterized queries, connection pooling, and transaction safety
- Monitoring Layer: Vercel Analytics, Speed Insights, and custom health checks
🎯 Context-Aware Security
Security requirements vary based on context. This portfolio balances public accessibility (portfolio sections) with protected resources (admin dashboard, MCP tools). Route-based middleware enables granular security policies matching the sensitivity of each endpoint.
📊 Implemented Security Metrics
Real-world security measurements from this production portfolio:
- Rate Limiting: Token bucket with 20 tokens/10s refill, 50 capacity per IP
- Database Performance: Connection pool max 20, 30s idle timeout, 10s connection timeout
- Cache Strategy: Public API routes with s-maxage=300, stale-while-revalidate=600
- Edge Runtime: experimental-edge for global middleware execution <50ms
- Vector Search: 1536-dimensional embeddings with cosine similarity for semantic security
📚 Documentation as Security Asset
Comprehensive documentation of security decisions, configurations, and rationale creates an audit trail, facilitates compliance, and serves as a learning resource. This security journal itself demonstrates the practice of documenting security implementations with specific metrics, lessons learned, and challenges overcome.
🚀 Production-Ready Security Stack
This portfolio showcases enterprise-grade security patterns suitable for production AI applications:
- ✅ Authentication & RBAC with Clerk (admin role in publicMetadata)
- ✅ Edge middleware with route-based protection (experimental-edge runtime)
- ✅ Application firewall with Arcjet (shield, bot detection, rate limiting)
- ✅ Secure database operations (parameterized queries, connection pooling, transactions)
- ✅ Environment variable management (separation of dev/staging/production)
- ✅ MCP protocol integration with security controls (Arcjet on /api/mcp)
- ✅ Performance monitoring (Vercel Analytics, Speed Insights, database health checks)
- 🔄 OAuth 2.1 for MCP servers (planned Week 7-8 implementation)